Loading ToolBench
AI-native code change analyst for intelligent changelog generation. Analyzes git commits, generates changelogs, detects breaking changes, and provides rich code context with tiered token budgeting.
ChangeWeave MCP has well-structured tool definitions with clear naming (verb_noun pattern) and comprehensive parameter schemas. All 7 tools include descriptions and typed input parameters with enums where appropriate. However, output schemas are not documented in the provided source, parameter descriptions lack depth (missing format/range constraints), and error handling guidance is absent. No tool annotations (readOnlyHint/destructiveHint) despite clear risk classifications. Descriptions average ~150 chars (within baseline 194), but lack actionable context for LLM selection.
Analyze commits between two refs with full diff context, categorization, and impact scoring. Returns detailed analysis including LOC stats, module mapping, and breaking change detection.
Compare two branches, tags, or commits to see all changes including file list, module mapping, and LOC statistics.
Analyze diffs between two refs to detect potential breaking changes: removed exports, changed function signatures, altered API routes, and schema migrations.
Generate a complete, publication-ready changelog from code changes between two refs. Provides rich context including diffs, categorization, and formatting instructions for different audiences and styles.
Get the full diff for a specific commit with file-level changes, hunks, categorization, and impact assessment.
Output schemas not documented. LLMs cannot plan downstream calls or extract required fields (e.g., does analyze_commits return breaking_changes array? What fields does each item contain?). Baseline: 100% of A+ tools document return types.
Parameter descriptions lack actionable constraints. E.g., 'depth' enum has no explanation of when to use 'summary' vs 'detailed' vs 'full'. 'max_commits' has no range guidance (1-1000? 1-10000?). LLMs cannot reason about valid values without explicit constraints.
No error handling guidance. Tools lack recovery hints (e.g., 'If commit not found, try compare_refs() to list available commits'). No categorization of errors as retryable vs user-fixable. Agents cannot self-correct on failures.
Inferred effective spec: 2026-07-28+.
| Scored | Grade | Overall | Spec posture | Rubric |
|---|---|---|---|---|
| 2026-09-23 | C | 62 | 2026-07-28+ | v2 |
List existing changelogs from the ChangeWeave platform for a repository.
Create and optionally publish a changelog to the ChangeWeave platform.
No tool annotations despite clear risk classifications. publish_changelog is WRITE (destructive), others are READ_ONLY. Missing destructiveHint/readOnlyHint annotations prevent agents from reasoning about side effects and retry safety.
GitHub token exposed as optional parameter. Per-request token override is a security anti-pattern, credentials should never appear in tool parameters. Use server-side secret injection via environment variables or vault.